ANALYSIS OF LINEAR REGRESSION AND SEMIVARIOGRAM MODELING (CASE: TOURISM FACTORS IN WEST SUMATERA PROVINCE)

• Main Author: HAUZAN FADLURRAHMAN (10113088), M.
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/23055
• Institution: Institut Teknologi Bandung

West Sumatera Province has a lot of natural and cultural wealth so that become a tourist destination. Tourism has several influencing factors such as the relationship between tourism factors and regency relationships that can be seen through linear regression analysis and isotropic semivariogram. Linear regression analysis can model the real regional income (PAD) in a regency/city against significant tourism factors and see the correlation between factors. The PAD regression model relies only on many tourism objects and tourism subsector income with adjusted coefficient of determination 89%. While the isotropic semivariogram illustrate the spatial relationship between the tourism condition of a regencies through the variance of the difference in observation value of a pair of locations that seperate by distance ℎ. In the data of each tourism factor, there are at least one very extreme value as outlier in Padang City. Therefore, in this case the Sturges approach is used in calculating the distribution of the number of distance lags to build experimental semivariogram. Some models can be matched with experimental semivariogram such as spherical, pentaspherical, and exponential. The best semivariogram model can be determinated by estimating three model parameters: nugget effect (𝐶0), partial sill (𝐶), and range (𝑎) that resulting the least of sum square error. The best semivariogram model is a pentaspherical model without 𝐶0, 𝐶 is from experimental semivariogram average, and 𝑎 is 45. This model means that the value of tourism in the capital of West Sumatera district has correlation with the maximum distance of 45 km and 17 vi pairs of correlated regency. Furthermore the model was used to estimate the large PAD in the new district through bootstrap on the ordinary kriging and Cholesky decomposition for correlated data. The bootstrap method is used because amount of tourism data in West Sumatera is small and contains an outliers. The tourism data from Cholesky decomposition is bootstrapped to obtain a large data size and calculated the average of the bootstrap results of each convergent location. The result of bootstrap kriging for PAD in a new regency that will be established is approximately 78-84 bilion rupiah.